On 8/18/2010 10:53 PM, Tobias Oetiker wrote: > > if you accquire your data at the sampling interval times then no > re-sampling will happen ... > > conside this > > * setup an rrd with a step of 300 seconds > this means it will save data at 0:00, 0:05, 0:10, 0:15, 0:20 > > * read data from your data-source. You get data at > 0:01 - 1, 0:05 - 33, 0:09 - 2, 0:15 - 8, 0:21 - 0 > > * what how would you like rrdtool to act ? what data should be > stored in the slots ? > > cheers > tobi > 1 33 2 8 0
You are thinking of it storing at very specific, to the nanosecond, instants. I am thinking of it storing in 5 minute windows (T +/- 2 minurtes 30 seconds). They are both problematic. There is virtually no way in the real world that a data source and rdd will by synchronized to the nanosecond. What happens if two values (or more) are presented within the same window? (you either drop the second value, or overwrite the previous one). But dividing the value by its position within the window makes no sense (for a gauge). Taking your own example of a car speedometer, that is the equivalent of assuming that the car comes to a stop and accelerates up to some speed once per sample period. Philip _______________________________________________ rrd-users mailing list [email protected] https://lists.oetiker.ch/cgi-bin/listinfo/rrd-users
